Predicting age of human subjects based on structural connectivity from diffusion tensor imaging

TL;DR

This study presents a straightforward method to predict individual human age from diffusion tensor imaging-based structural connectivity, achieving a strong correlation with actual age in a diverse age group.

Contribution

The paper introduces a simple, connectivity-based age prediction method using DTI data, emphasizing individual assessment over group differences.

Findings

Strong correlation (rho=0.77) between predicted and actual age.

Average deviation of 10 years in age prediction.

Effective across a wide age range (4-85 years).

Abstract

Predicting brain maturity using noninvasive magnetic resonance images (MRI) can distinguish different age groups and help to assess neurodevelopmental disorders. However, group-wise differences are often less informative for assessing features of individuals. Here, we propose a simple method to predict the age of an individual subject solely based on structural connectivity data from diffusion tensor imaging (DTI). Our simple predictor computed a weighted sum of the strength of all connections of an individual. The weight consists of the fiber strength, given by the number of streamlines following tract tracing, multiplied by the importance of that connection for an observed feature--age in this case. We tested this approach using DTI data from 121 healthy subjects aged 4 to 85 years. After determining importance in a training dataset, our predicted ages in the test dataset showed a strong correlation (rho = 0.77) with real age deviating by, on average, only 10 years.